What you get back
Short paragraphs, each marked with the second it starts at. Click a line to jump to it,
copy a quote with its timecode, or export the lot. .txt keeps the timings
inline; .srt and .vtt give you the same words as a subtitle file.
We tested it on a public SoundCloud podcast episode: a 10-minute-50-second interview that opens with a short ad and a film clip before the host and a psychologist take over. The audio arrived in 3 seconds; the free lane then took 12 minutes, so from pasting the link to finished text was 12 minutes 20 seconds. It came back as 149 timestamped lines, language detected as English. The stretch we read line by line was clean, film clip included. The slip we caught was a name: the guest's surname, printed right there in the track title, came out spelled the way it sounds. Check names before you quote them.
The engine is OpenAI Whisper, large-v3-turbo. If the track is not in English, read how accurate Whisper actually is before you lean on the text. The gap between languages is wider than most people expect.
Which links work, and which can't
- One track: yes.
soundcloud.com/artist/track, or the shorton.soundcloud.comlink from the Share button in the app. Both are accepted. - A private track with its secret link: yes. If the link plays for someone who is not logged in, we can fetch it the same way.
- A set or a profile: no. Addresses with
/sets/in them, and artist pages, are lists. Open the track you want and paste that link instead. - Preview-only tracks: 30 seconds. Some tracks play only a 30-second preview unless you subscribe. We checked one: its page lists a 27-minute recording, and the address hands out exactly 30 seconds of it. Without a subscriber's login, that is all a link can get.
What people actually send us from SoundCloud
Mostly talk, by the look of it. Between September 10 and October 20, 2026 visitors ran 40 SoundCloud tracks through shrybe. Half were under eight minutes, and 17 ran past a quarter of an hour, up to a full hour. They came in six languages: English for most, Spanish for ten, and one or two each of Italian, Turkish, Norwegian and Estonian. Where the address shows the track's name, 8 of 20 have "podcast", "episode" or "show" in it, and none has "mix", "remix" or "feat" – the words that usually mark a song.
That fits how the platform is used. A lot of spoken-word audio ends up there because uploading is free and the player embeds anywhere, but a track page carries no transcript, so the words stay locked inside the player. Turned into text, an episode can be searched, quoted exactly, read by someone who cannot listen, or turned into .srt or .vtt captions for a video cut of the same show.
Music versus speech
Whisper was trained to transcribe speech, and that is where it is good. On a song it tries to write down the vocals, but a full mix buries the words, so the result is patchy. Stretches without speech are the other weak spot. A 2024 study of Whisper by Koenecke and colleagues at Cornell (ACM FAccT) found entire invented phrases in roughly 1% of the transcriptions it checked, more often where the speaker paused for long. That was an earlier version of the model, and the authors say the rate has dropped since, but the pattern is worth knowing. We would not rely on it for lyrics. For a DJ mix, expect little more than the talking between tracks. A talk with quiet music under it usually goes fine, because the voice sits on top.
Long tracks and waiting times
The free tier stops at 75 minutes of audio per transcript, fifteen transcripts a day and 180 minutes of audio a day from one address. Most SoundCloud talks fit inside that. A long radio show or a recorded conference session may not, and that is what the paid fast lane is for: files up to three hours, $5 for 10 hours of audio.
The free lane works at about real time, so a 60-minute episode takes roughly an hour, and a busy hour adds a queue that the page shows you as it moves. The fast lane runs the same model about 69 times faster. Leave the tab open on the free lane; the transcript appears when it is done.
When a link won't go through
Three usual causes. The track was deleted or made private. The link is a set or a profile rather than a track. Or the recording is longer than the free 75 minutes. If a track that plays in a private browser window still fails, try it once more before deciding the link is the problem.
Have the file itself? Upload it or paste a direct link instead. Podcasts that also have an RSS feed go through podcast transcription, and a recording that is not in English can come back translated. There are pages for YouTube and TikTok too.
Frequently asked
Which SoundCloud links work?
A link to one track. That is the address with the artist and the track name in it (soundcloud.com/artist/track), or the short on.soundcloud.com link the app gives you from the Share button. A private track shared with its secret link works too, as long as the link opens for someone who is not logged in.
Can I paste a playlist or a whole profile?
No. A set (an address with /sets/ in it) or a profile page is a list of tracks, and each track needs its own link. Paste up to three track links at once and they run as a batch.
Why did I only get 30 seconds of text?
Some tracks are set to play only a 30-second preview to anyone without a SoundCloud subscription. The track page then hands out those 30 seconds and nothing more, so that is all a link can give us. There is no way around it without the creator's file: if you have it, upload it instead (up to 100 MB).
Does it transcribe song lyrics?
It writes down what it hears, and it was built for speech. On a podcast, an interview or a lecture that is exactly what you want. On a song, sung words over a full mix come back patchy at best, and a purely instrumental track has nothing to write down.
Do I need a SoundCloud account?
No account, and nothing to install. If a track plays in a browser for someone who is not logged in, it can be transcribed. A track that has been deleted or made private without a secret link cannot, and you get an error instead of a transcript.
How long can a track be?
On the free tier, up to 75 minutes of audio per transcript. Longer recordings, such as a full radio show or a three-hour DJ set with talk between the tracks, need the paid fast lane, which takes files up to three hours: $5 for 10 hours of audio.
Can I get subtitles rather than plain text?
Yes. Every transcript can be saved as .txt, .srt or .vtt, and the two subtitle formats carry the same timings as the text.
Copy the track link, paste it above.
Need subtitles rather than notes? The caption export is the same transcript in a different wrapper.